MCP Coursework - Model Context Protocol Learning Project

This repository contains my coursework and learning materials for the Model Context Protocol (MCP), including server implementations, client applications, and practical examples.

📁 Project Structure

mcp_coursework/
├── mcp_project/                 # Main project directory
│   ├── research_server.py       # MCP server for arXiv paper research
│   ├── mcp_chatbot.py          # MCP client chatbot implementation
│   ├── server_config.json      # Multi-server configuration
│   ├── pyproject.toml          # Python project configuration
│   └── papers/                 # Research papers data (gitignored)
├── mcp_client.ipynb            # Single server client notebook
├── mcp_client_multi_server.ipynb # Multi-server client notebook
├── research_server.ipynb       # Server creation tutorial notebook
└── .gitignore                  # Git ignore rules

🚀 Features

MCP Research Server

• arXiv Paper Search: Search and retrieve academic papers from arXiv
• Paper Information Storage: Save paper metadata locally
• Information Extraction: Retrieve stored paper information by ID

MCP Client Applications

• Single Server Chatbot: Connect to one MCP server
• Multi-Server Chatbot: Connect to multiple MCP servers simultaneously
• Tool Discovery: Automatic tool schema discovery and registration

Supported MCP Servers

• Research Server: Custom arXiv paper research tools
• Filesystem Server: File and directory operations
• Fetch Server: Web content retrieval

🛠️ Setup Instructions

Prerequisites

• Python 3.13+
• uv package manager
• Anthropic API key (for Claude models)

Installation

1. Clone the repository

   git clone https://github.com/alicegu1117/MCP_starter.git
   cd MCP_starter

2. Set up the Python environment

   cd mcp_project
   uv venv
   source .venv/bin/activate  # On Windows: .venv\Scripts\activate
   uv sync

3. Configure API keys Create a .env file in the mcp_project directory:

   ANTHROPIC_API_KEY=your_anthropic_api_key_here

📖 Usage Examples

Running the Research Server

cd mcp_project
uv run research_server.py

Running the Multi-Server Chatbot

cd mcp_project
python mcp_chatbot.py

Using Jupyter Notebooks

cd mcp_project
jupyter notebook
# Open mcp_client.ipynb or mcp_client_multi_server.ipynb

🔧 Key Components

Research Server (research_server.py)

• search_papers(topic, max_results): Search arXiv for papers on a topic
• extract_info(paper_id): Retrieve stored paper information

MCP Chatbot (mcp_chatbot.py)

• Connects to multiple MCP servers
• Automatic tool discovery and registration
• LLM-powered conversation with tool usage

Server Configuration (server_config.json)

{
    "mcpServers": {
        "research": {
            "command": "uv",
            "args": ["run", "research_server.py"]
        },
        "filesystem": {
            "command": "npx",
            "args": ["-y", "@modelcontextprotocol/server-filesystem", "."]
        }
    }
}

📚 Learning Materials

This project includes several Jupyter notebooks that demonstrate:

• Lesson 3: Basic MCP client implementation
• Lesson 4: Creating an MCP server with FastMCP
• Lesson 6: Multi-server MCP client implementation

🤝 Contributing

This is a learning project, but suggestions and improvements are welcome!

📄 License

This project is for educational purposes as part of MCP coursework.

🔗 Resources

• Model Context Protocol Documentation
• FastMCP Library
• MCP Python SDK
• MCP Inspector

---

Note: The papers/ directory contains generated research data and is excluded from version control via .gitignore.